Attribution
Part of Gaming campaign attribution
Reconciling network outcomes with first-party events
Match event definitions, populations, dates and counting rules to explain gaps between gaming network reports and app records.
To reconcile a network outcome with an app or transaction event, align both sides on the same action, population and period.
Start with event counts and definitions. A network's attributed total can differ from all app events: it credits only eligible ad interactions, or the systems count and date actions differently.
Build a comparison row
For each network conversion action, identify the closest app or transaction event. Record the trigger, app and platform, population, counting unit, date basis, time zone, attribution window and known exclusions.
If the events are merely related, label the row a related comparison rather than a reconciliation.
| Field | Network side | App or transaction side |
|---|---|---|
| Event | Credited conversion action and trigger | Observed event and trigger |
| Population | Eligible attributed activity | All observed activity or a supported matched cohort |
| Date | Interaction or conversion date, as selected | Event or transaction date |
| Counting | One or every qualifying action | Events, unique users or transactions |
| Coverage | Imports, platforms and attribution limits | Instrumented versions, payment routes and missing events |
Do not divide all app purchases by attributed network installs and call the result an attributed purchase rate. Without supported cohort linkage, the populations overlap only partly.
Work through the difference
First check the event. Play installs and Firebase first opens are different actions.
For purchases, confirm whether an event represents a completed transaction or only an attempted action. Where records exist, inspect payment status and duplicate handling.
Next check the network conversion action. Confirm whether the event was imported for the period and whether the selected report includes it. Check the relevant configuration to see how the action is counted and displayed.
Check the counting option and the applicable click, engaged-view, view-through or post-install windows. Apply the same questions to another seller only after checking that seller's definitions.
Then align dates and time zones. An outcome organised by ad-interaction date can sit in a different week from the app event recorded on its occurrence date. Allow the outcome period to mature and retain each original extract with its export time.
Finally check instrumentation and coverage. For an app using Firebase, DebugView can help a developer inspect development-device events and parameters.
Record app versions, devices and payment routes for which events are unavailable. Leave unmeasured activity unknown rather than allocating it to a campaign.
Steps to Reconcile Network Outcomes with First-Party Events
- Verify the event definitionConfirm whether actions like 'play install' and 'Firebase first open' are distinct.
- Validate purchase statusCheck if the event reflects a completed transaction or an attempt; inspect payment status and duplicates.
- Confirm network conversion configurationEnsure the event was imported and included in the report; review counting and display settings.
- Align counting rules and windowsCheck click, engaged-view, view-through or post-install attribution windows for consistency.
- Harmonise dates and time zonesAccount for differences between ad-interaction date and app event occurrence date; retain original extracts.
- Assess instrumentation and coverageUse Firebase DebugView to inspect device-level events; record app versions, devices and payment routes with missing data.
Explain the remaining gap
Classify supported differences as event definition, attribution or population, timing, and unresolved data. Assign an amount to a category only when the records support it. If a material residual remains, obtain the relevant configuration and event-delivery evidence before using the figure to shift spend.
Show both original totals, any genuinely aligned comparison and the unexplained remainder. Agreement between systems does not establish incremental effect; reconciliation establishes whether their observed figures are compatible.
